def df_contour(df,
xcol,
ycol,
zcol,
df_min,
lim_params,
correct=None,
logscale=False,
dof=1):
df_lim = df.copy()
for param in lim_params:
df_lim = df_lim[df_lim[param] == df_min[param].values[0]]
dfpivot = df_lim.pivot(xcol, ycol, zcol)
Y = dfpivot.columns.values
X = dfpivot.index.values
Z = dfpivot.values
print(X, Y, Z.shape)
x, y = np.meshgrid(X, Y, indexing="ij")
fig, ax = plt.subplots()
if logscale:
c = ax.contourf(x, y, Z, locator=ticker.LogLocator(), cmap=cm.viridis)
else:
c = ax.contourf(x, y, Z, cmap=cm.viridis)
# Chi levels values
print("Using chisquare dof=", dof)
sigmas = [
Z.ravel()[Z.argmin()] + chi2_at_sigma(sigma, dof=dof)
for sigma in range(1, 6)
]
sigma_labels = {
sigmas[sig - 1]: "${}-\sigma$".format(sig)
for sig in range(1, 6)
}
c2 = plt.contour(c, levels=sigmas)
plt.clabel(c2, fmt=sigma_labels, colors='w', fontsize=14)
cbar = plt.colorbar(c)
cbar.ax.set_ylabel(zcol)
plt.xlabel(xcol)
plt.ylabel(ycol)
if correct:
# Correct location of simulation
plt.plot(correct[xcol], correct[ycol], "ro", markersize=7)
# Mark minimum with a +.
min_i, min_j = divmod(Z.argmin(), Z.shape[1])
plt.plot(X[min_i], Y[min_j], "g*", markersize=7, label="$Min \chi^2$")
plt.show()
def chi2_individual_parabola_plots(table, params):
extractor = DBExtractor(table)
parabola_list = ["teff_1", "logg_1", "feh_1", "gamma"]
for par in parabola_list:
df = extractor.simple_extraction(columns=[par])
unique_par = list(set(df[par].values))
unique_par.sort()
for chi2_val, npix_val in zip(chi2_names, npix_names):
min_chi2 = []
for unique_val in unique_par:
df_chi2 = extractor.fixed_ordered_extraction(
columns=[par, chi2_val],
order_by=chi2_val,
limit=3,
fixed={par: float(unique_val)},
asc=True)
min_chi2.append(df_chi2[chi2_val].values[0])
# min_chi2 = reduced_chi_squared(min_chi2, params["npix"][npix_val], params["npars"])
min_chi2 = min_chi2 - min(min_chi2)
plt.plot(unique_par, min_chi2, ".-", label=chi2_val)
# popt, _ = curve_fit(parabola, unique_par, min_chi2)
popt, _ = fit_chi2_parabola(unique_par, min_chi2)
# print("params", popt)
x = np.linspace(unique_par[0], unique_par[-1], 40)
plt.plot(x, parabola(x, *popt)) # , label="parabola")
plt.xlabel(r"${0}$".format(par))
plt.ylabel(r"$\Delta \chi^2$ from mimimum")
plt.ylim([-0.05 * np.max(min_chi2), np.max(min_chi2)])
# Find roots
try:
residual = lambda x: parabola(x, *popt) - chi2_at_sigma(
1, params["npars"])
min_chi2_par = unique_par[np.argmin(min_chi2)]
min_chi2_par.astype(np.float64)
try:
lower_bound = newton(
residual,
(min_chi2_par + unique_par[0]) / 2) - min_chi2_par
except RuntimeError as e:
print(e)
lower_bound = np.nan
try:
upper_bound = newton(
residual,
(min_chi2_par + unique_par[-1]) / 2) - min_chi2_par
except RuntimeError as e:
print(e)
upper_bound = np.nan
print("min_chi2_par", min_chi2_par,
type(min_chi2_par), "\nlower_bound", lower_bound,
type(lower_bound), "\nupper_bound", upper_bound,
type(upper_bound))
print("{0} solution {1: 5.3} {2:+5.3} {3:+5.3}".format(
chi2_val, float(min_chi2_par), float(lower_bound),
float(upper_bound)))
plt.annotate("{0: 5.3f} {1:+5.3f} {2:+5.3f}".format(
float(min_chi2_par), (lower_bound), (upper_bound)),
xy=(float(min_chi2_par), 0),
xytext=(0.4, 0.5),
textcoords="figure fraction",
arrowprops={"arrowstyle": "->"})
except Exception as e:
print(e)
logging.warning("Could not Annotate the contour plot")
plt.axhline(y=chi2_at_sigma(1, params["npars"]), label="1 sigma")
plt.axhline(y=chi2_at_sigma(2, params["npars"]), label="2 sigma")
plt.axhline(y=chi2_at_sigma(3, params["npars"]), label="3 sigma")
plt.legend()
filename = "Chi2_Parabola_fit_{0}-{1}_{2}_param_{3}_{4}_individual_{5}.png".format(
params["star"], params["obsnum"], params["chip"], par,
params["suffix"], chi2_val)
plt.savefig(os.path.join(params["path"], "plots", filename))
plt.close()
print("saved individual chi2 parabolas for ", par)
plt.close()